@barivia/barsom-mcp 0.22.1 → 0.22.4

package/dist/tools/train.js

@@ -1,7 +1,8 @@
 import { z } from "zod";
 import { registerAuditedTool } from "../audit.js";
-import { apiCall, textResult } from "../shared.js";
+import { apiCall } from "../shared.js";
 import { buildTrainSubmitExtras } from "../train_submit_message.js";
+import { TRAINING_MONITOR_URI, buildTrainingMonitorResult, } from "./training_monitor.js";
 import { TRAINING_INPUT_SCHEMA, buildTrainMapParams, buildFloopParams, fetchTrainingPresets, assertValidNormalizationArgs, } from "./training_core.js";
 export const TRAIN_DESCRIPTION = `Submit a self-organizing map training job. Returns a job_id; poll with jobs(action=status) and then results(action=get). All actions are async.
 
@@ -89,7 +90,12 @@ async function submitTrainJob(opts) {
     }
     data.message = msg;
     data.suggested_next_step = extras.suggested_next_step;
-    return textResult(data);
+    const monitor = await buildTrainingMonitorResult(newJobId, { preamble: msg });
+    return {
+        content: [{ type: "text", text: JSON.stringify(data, null, 2) }, ...monitor.content],
+        structuredContent: monitor.structuredContent,
+        _meta: { ui: { resourceUri: TRAINING_MONITOR_URI } },
+    };
 }
 /**
  * Core training dispatcher. Shared by the `train` tool and the deprecated
@@ -98,7 +104,7 @@ async function submitTrainJob(opts) {
 export async function runTrain(action, args) {
     const dataset_id = args.dataset_id;
     if (action === "map" || action === "siom_map" || action === "impute") {
-        const { preset, grid_x, grid_y, epochs, model, periodic, columns, cyclic_features, cyclic_pairs, temporal_features, feature_weights, transforms, auto_log_transforms, time_delay_embeddings, categorical_features, normalize, normalization_methods, sigma_f, learning_rate, batch_size, quality_metrics, backend, output_format, output_dpi, colormap, row_range, gamma, gamma_f, siom_decay, siom_penalty, penalty_alpha, reset_per_epoch, siom_feature_geometry, siom_qe_backend, siom_qe_batch_size, label, cv_folds, viz_mode, viz_top_components, emit_cell_uncertainty, } = args;
+        const { preset, grid_x, grid_y, epochs, model, periodic, columns, cyclic_features, cyclic_pairs, temporal_features, feature_weights, transforms, auto_log_transforms, time_delay_embeddings, categorical_features, normalize, normalization_methods, sigma_f, learning_rate, batch_size, quality_metrics, backend, output_format, output_dpi, colormap, row_range, gamma, gamma_f, siom_decay, siom_penalty, penalty_alpha, reset_per_epoch, siom_feature_geometry, siom_qe_backend, siom_qe_batch_size, label, cv_folds, viz_mode, viz_top_components, viz_upload_cyclic_components, emit_cell_uncertainty, } = args;
         let PRESETS = {};
         try {
             PRESETS = await fetchTrainingPresets();
@@ -123,7 +129,8 @@ export async function runTrain(action, args) {
             temporal_features, feature_weights, transforms, auto_log_transforms,
             time_delay_embeddings, categorical_features,
             normalize, normalization_methods, sigma_f, learning_rate, batch_size, quality_metrics, backend,
-            output_format, output_dpi, colormap, row_range,
+            output_format, output_dpi, colormap, row_range, siom_feature_geometry, siom_qe_backend, siom_qe_batch_size,
+            viz_upload_cyclic_components,
         }, PRESETS);
         if (action === "impute") {
             params._job_type = "train_impute";

package/dist/tools/training_core.js

@@ -113,6 +113,8 @@ export const TRAINING_INPUT_SCHEMA = {
         .describe("Visualization density; auto-capped when feature count > 40"),
     viz_top_components: z.number().int().min(0).max(64).optional()
         .describe("With viz_mode=summary_plus_top: upload top-N component maps by variance (default 8)"),
+    viz_upload_cyclic_components: z.boolean().optional()
+        .describe("When true with individual component PNGs, also publish cos/sin cyclic expansion planes (default false; recovered hour/month panels stay in combined.png)"),
     emit_cell_uncertainty: z.boolean().optional()
         .describe("impute: write imputation_uncertainty.csv (pool_std = prototype-neighbourhood dispersion, not calibrated SD; pool_n = contributing nodes)"),
     gamma: z.preprocess((v) => (v !== undefined && v !== null && typeof v === "string") ? parseFloat(v) : v, z.number().optional())
@@ -245,7 +247,7 @@ export function buildFloopParams(args) {
     return { params, paramSummary };
 }
 export function buildTrainMapParams(args, presets) {
-    const { preset, grid_x, grid_y, epochs, model, periodic, columns, cyclic_features, cyclic_pairs, temporal_features, feature_weights, transforms, auto_log_transforms, time_delay_embeddings, categorical_features, normalize, normalization_methods, sigma_f, learning_rate, batch_size, quality_metrics, backend, output_format, output_dpi, colormap, row_range, siom_feature_geometry, siom_qe_backend, siom_qe_batch_size, } = args;
+    const { preset, grid_x, grid_y, epochs, model, periodic, columns, cyclic_features, cyclic_pairs, temporal_features, feature_weights, transforms, auto_log_transforms, time_delay_embeddings, categorical_features, normalize, normalization_methods, sigma_f, learning_rate, batch_size, quality_metrics, backend, output_format, output_dpi, colormap, row_range, siom_feature_geometry, siom_qe_backend, siom_qe_batch_size, viz_upload_cyclic_components, } = args;
     const p = preset ? presets[preset] : undefined;
     const params = {
         model: model ?? "SOM",
@@ -308,6 +310,9 @@ export function buildTrainMapParams(args, presets) {
         params.siom_qe_backend = siom_qe_backend;
     if (siom_qe_batch_size !== undefined)
         params.siom_qe_batch_size = siom_qe_batch_size;
+    if (viz_upload_cyclic_components !== undefined) {
+        params.viz_upload_cyclic_components = viz_upload_cyclic_components;
+    }
     const effectiveGrid = params.grid;
     const effectiveEpochs = params.epochs;
     const effectiveBatch = params.batch_size;

package/dist/tools/training_monitor.js

@@ -1,8 +1,10 @@
 import { z } from "zod";
 import { registerAppTool } from "@modelcontextprotocol/ext-apps/server";
 import { runMcpToolAudit } from "../audit.js";
-import { apiCall, getClientSupportsMcpApps, getVizPort, structuredTextResult, } from "../shared.js";
+import { apiCall, BARSOM_VIZ_VIEWS, buildStandaloneVizPageUrl, buildVizRelatedUrls, getVizPort, structuredTextResult, } from "../shared.js";
+import { appendUiDeliveryContent, beginInlineFallback, resolveUiDelivery, tryAttachLearningCurve, } from "../ui-delivery.js";
 import { lastEpochTeFromCurves } from "../training_monitor_curve.js";
+import { buildVizStandaloneUrl, resultsExplorerStandaloneLinkText, } from "../viz_links.js";
 export const TRAINING_MONITOR_URI = "ui://barsom/training-monitor";
 export const TRAINING_MONITOR_REFRESH_MS = 5000;
 function hasCurveArrays(data) {
@@ -14,8 +16,12 @@ function hasCurveArrays(data) {
     ];
     return keys.some((k) => Array.isArray(data[k]) && data[k].length > 0);
 }
-function buildStructuredContent(job_id, data) {
+function buildStructuredContent(job_id, data, port = getVizPort()) {
     const id = String(data.id ?? job_id);
+    const standaloneUrl = buildStandaloneVizPageUrl(port, BARSOM_VIZ_VIEWS.trainingMonitor, id) ??
+        buildVizStandaloneUrl("training-monitor", id);
+    const resultsExplorerUrl = buildStandaloneVizPageUrl(port, BARSOM_VIZ_VIEWS.resultsExplorer, id) ??
+        buildVizStandaloneUrl("results-explorer", id);
     return {
         ...data,
         type: "training-monitor",
@@ -23,6 +29,9 @@ function buildStructuredContent(job_id, data) {
         id,
         job_id: id,
         refresh_interval_ms: TRAINING_MONITOR_REFRESH_MS,
+        standaloneUrl,
+        resultsExplorerUrl,
+        relatedUrls: buildVizRelatedUrls(port, id, BARSOM_VIZ_VIEWS),
     };
 }
 function hasTeCurveArrays(data) {
@@ -81,6 +90,71 @@ async function enrichWithTrainingLog(job_id, data) {
         return data;
     }
 }
+export async function buildTrainingMonitorResult(job_id, opts) {
+    const { fetch_training_log, preamble } = opts ?? {};
+    let data = (await apiCall("GET", `/v1/jobs/${job_id}`));
+    const jobStatus = String(data.status ?? "");
+    if (fetch_training_log || needsTrainingLogEnrichment(data)) {
+        data = await enrichWithTrainingLog(job_id, data);
+    }
+    const port = getVizPort();
+    const structuredContent = buildStructuredContent(job_id, data, port);
+    const status = jobStatus;
+    const progress = (data.progress ?? 0) * 100;
+    const etaSec = data.training_eta_sec != null ? Number(data.training_eta_sec) : null;
+    const elapsedSec = data.training_elapsed_sec != null ? Number(data.training_elapsed_sec) : null;
+    const epoch = data.epoch != null ? Number(data.epoch) : null;
+    const totalEpochs = data.total_epochs != null ? Number(data.total_epochs) : null;
+    const timingParts = [];
+    if (elapsedSec != null && elapsedSec >= 0)
+        timingParts.push(`elapsed ${Math.round(elapsedSec)}s`);
+    if (etaSec != null && etaSec > 0)
+        timingParts.push(`ETA ~${Math.round(etaSec)}s`);
+    if (epoch != null && totalEpochs != null)
+        timingParts.push(`epoch ${epoch}/${totalEpochs}`);
+    const timingNote = timingParts.length > 0 ? ` ${timingParts.join(", ")}.` : "";
+    const text = (preamble ? `${preamble}\n\n` : "") +
+        `Training monitor (live when embedded, refreshes every ${TRAINING_MONITOR_REFRESH_MS / 1000}s): job ${job_id} — ${status} (${progress.toFixed(1)}%).${timingNote} ` +
+        `Optional UI; jobs(action=status) is enough to finish the workflow headless.`;
+    const content = [{ type: "text", text }];
+    const terminal = status === "completed" || status === "failed" || status === "cancelled";
+    let inlineImageAttached = false;
+    if (terminal) {
+        inlineImageAttached = await tryAttachLearningCurve(content, job_id);
+    }
+    const uiDelivery = resolveUiDelivery({
+        tool: "training_monitor",
+        jobId: job_id,
+        jobStatus: status,
+        inlineImageAttached,
+    });
+    appendUiDeliveryContent(content, uiDelivery, {
+        linkLabel: "Open training monitor",
+        primaryUrl: structuredContent.standaloneUrl,
+    });
+    if (status === "completed" && uiDelivery.urls?.length) {
+        const resultsUrl = uiDelivery.urls.find((u) => u.includes("results-explorer"));
+        if (resultsUrl) {
+            content.push({
+                type: "text",
+                text: `Job completed — browse map figures: [Open results explorer](${resultsUrl})`,
+            });
+        }
+    }
+    structuredContent.ui_delivery = uiDelivery;
+    return { structuredContent, content, text };
+}
+/** After terminal + finalize, append results explorer open hint (monitor completion path). */
+export function appendResultsExplorerCompletionHint(content, job_id) {
+    const link = resultsExplorerStandaloneLinkText(job_id);
+    if (link) {
+        content.push({
+            type: "text",
+            text: `${link}\n` +
+                `AGENT: when the job is fully complete (finalize done if any), call results_explorer(job_id="${job_id}") in the same turn and surface this link — do not wait for the user to ask.`,
+        });
+    }
+}
 export function registerTrainingMonitorTool(server) {
     registerAppTool(server, "training_monitor", {
         title: "Training Monitor",
@@ -94,42 +168,11 @@ export function registerTrainingMonitorTool(server) {
         },
         _meta: { ui: { resourceUri: TRAINING_MONITOR_URI } },
     }, async (args) => runMcpToolAudit("training_monitor", "default", args, async () => {
+        beginInlineFallback();
         const { job_id, fetch_training_log } = args;
-        let data = (await apiCall("GET", `/v1/jobs/${job_id}`));
-        const jobStatus = String(data.status ?? "");
-        if (fetch_training_log || needsTrainingLogEnrichment(data)) {
-            data = await enrichWithTrainingLog(job_id, data);
-        }
-        const structuredContent = buildStructuredContent(job_id, data);
-        const status = jobStatus;
-        const progress = (data.progress ?? 0) * 100;
-        const etaSec = data.training_eta_sec != null ? Number(data.training_eta_sec) : null;
-        const elapsedSec = data.training_elapsed_sec != null ? Number(data.training_elapsed_sec) : null;
-        const epoch = data.epoch != null ? Number(data.epoch) : null;
-        const totalEpochs = data.total_epochs != null ? Number(data.total_epochs) : null;
-        const timingParts = [];
-        if (elapsedSec != null && elapsedSec >= 0)
-            timingParts.push(`elapsed ${Math.round(elapsedSec)}s`);
-        if (etaSec != null && etaSec > 0)
-            timingParts.push(`ETA ~${Math.round(etaSec)}s`);
-        if (epoch != null && totalEpochs != null)
-            timingParts.push(`epoch ${epoch}/${totalEpochs}`);
-        const timingNote = timingParts.length > 0 ? ` ${timingParts.join(", ")}.` : "";
-        const text = `Training monitor (live when embedded, refreshes every ${TRAINING_MONITOR_REFRESH_MS / 1000}s): job ${job_id} — ${status} (${progress.toFixed(1)}%).${timingNote} ` +
-            `Optional UI; jobs(action=status) is enough to finish the workflow.`;
-        const content = [{ type: "text", text }];
-        const port = getVizPort();
-        const standaloneUrl = port
-            ? `http://localhost:${port}/viz/training-monitor?mode=standalone&job_id=${encodeURIComponent(job_id)}`
-            : undefined;
-        if (!getClientSupportsMcpApps() && standaloneUrl) {
-            content.push({
-                type: "text",
-                text: `Live training monitor: [Open training monitor](${standaloneUrl})\n` +
-                    `AGENT: surface this to the user as a clickable markdown link (e.g. "[Open training monitor](${standaloneUrl})") in your reply — do not leave it buried in tool output. Re-post the link after the job completes so the user can return to the curves.\n` +
-                    `This localhost port is assigned per MCP session and changes if the proxy restarts — if the page stays at "running 0%" or stops updating, the link is stale: re-run training_monitor for a fresh URL, or set BARIVIA_VIZ_PORT in your MCP env for a persistent port. Health check: http://localhost:${port}/api/health?job_id=${encodeURIComponent(job_id)}`,
-            });
-        }
+        const { structuredContent, content, text } = await buildTrainingMonitorResult(job_id, {
+            fetch_training_log,
+        });
         return {
             ...structuredTextResult(structuredContent, text, content),
             _meta: { ui: { resourceUri: TRAINING_MONITOR_URI } },

package/dist/train_submit_message.js

@@ -1,17 +1,15 @@
-import { getVizPort } from "./shared.js";
+import { buildVizStandaloneUrl } from "./viz_links.js";
 export function buildTrainSubmitExtras(opts) {
     const { newJobId, totalRows, hasTransforms, prepareJobId, variantPrefix, paramSummary, impute, resultsHint } = opts;
-    const monitorUrl = getVizPort() > 0
-        ? `http://localhost:${getVizPort()}/viz/training-monitor?mode=standalone&job_id=${encodeURIComponent(newJobId)}`
-        : null;
+    const monitorUrl = buildVizStandaloneUrl("training-monitor", newJobId);
     let msg = `Job submitted (${variantPrefix}, ${paramSummary}). `;
     if (prepareJobId) {
         msg += `Preprocessing job prepare_training_matrix (${prepareJobId}) runs on worker-io first; train job ${newJobId} starts after it completes. Poll jobs(action=status, job_id="${prepareJobId}") until completed, then poll the train job. `;
     }
-    msg += `Recommended: training_monitor(job_id="${newJobId}")`;
+    msg += `Training monitor opened for job ${newJobId}`;
     if (monitorUrl)
-        msg += ` or open ${monitorUrl}`;
-    msg += ` while polling jobs(action=status, job_id="${newJobId}") every 60–120s for large jobs. `;
+        msg += ` — standalone: ${monitorUrl}`;
+    msg += `. Poll jobs(action=status, job_id="${newJobId}") every 60–120s for large jobs if needed. `;
     msg += `When status is completed, check jobs(action=status) for finalize_job_id — if present, poll that finalize_training job before results(action=get). `;
     msg += `Then use results(action=get, job_id="${newJobId}")`;
     if (resultsHint) {

package/dist/training_monitor_curve.js

@@ -83,3 +83,35 @@ export function lastEpochTeFromCurves(data) {
     }
     return null;
 }
+export function isTerminalJobStatus(status) {
+    return status === "completed" || status === "failed" || status === "cancelled";
+}
+/** True when compute + finalize (if any) are done — safe to open results_explorer. */
+export function isFullyComplete(data) {
+    const status = String(data.status ?? "");
+    if (status !== "completed")
+        return isTerminalJobStatus(status);
+    const finalizeId = data.finalize_job_id != null && String(data.finalize_job_id) !== ""
+        ? String(data.finalize_job_id)
+        : null;
+    if (!finalizeId)
+        return true;
+    return String(data.finalize_status ?? "") === "completed";
+}
+/** Keep polling while finalize is still running after parent compute completed. */
+export function shouldKeepPollingJob(data) {
+    const status = String(data.status ?? "");
+    if (!isTerminalJobStatus(status))
+        return true;
+    if (status !== "completed")
+        return false;
+    if (data.finalize_failed === true)
+        return false;
+    const finalizeId = data.finalize_job_id != null && String(data.finalize_job_id) !== ""
+        ? String(data.finalize_job_id)
+        : null;
+    const finalizeStatus = data.finalize_status != null && String(data.finalize_status) !== ""
+        ? String(data.finalize_status)
+        : null;
+    return finalizeId != null && finalizeStatus !== "completed";
+}

package/dist/ui-delivery.js

@@ -0,0 +1,185 @@
+/**
+ * Client-capability adaptation for MCP App tools (training_monitor, results_explorer).
+ * Respects BARIVIA_UI_DELIVERY override and emits a structured block agents can rely on.
+ */
+import { apiRawCall, getClientSupportsMcpApps, getVizPort, mimeForFilename, resetInlineAttachBudget, } from "./shared.js";
+import { buildVizStandaloneUrl } from "./viz_links.js";
+export function parseUiDeliveryOverride() {
+    const raw = (process.env.BARIVIA_UI_DELIVERY ?? "auto").trim().toLowerCase();
+    if (raw === "apps" || raw === "localhost" || raw === "inline" || raw === "auto") {
+        return raw;
+    }
+    return "auto";
+}
+function vizPortPinned() {
+    return Boolean(process.env.BARIVIA_VIZ_PORT?.trim());
+}
+function viewForTool(tool) {
+    return tool === "training_monitor" ? "training-monitor" : "results-explorer";
+}
+/** Prefer viz_links (single URL builder) so train/monitor/explorer stay consistent. */
+export function buildStandaloneVizUrl(view, jobId) {
+    return buildVizStandaloneUrl(viewForTool(view), jobId);
+}
+export function buildHealthUrl(jobId) {
+    const port = getVizPort();
+    if (!port)
+        return undefined;
+    const base = `http://localhost:${port}/api/health`;
+    return jobId ? `${base}?job_id=${encodeURIComponent(jobId)}` : base;
+}
+function hintForTier(tier, tool, opts) {
+    const label = tool === "training_monitor" ? "training monitor" : "results explorer";
+    switch (tier) {
+        case "embedded":
+            return `The ${label} MCP App panel is embedded in this client. Also mention the standalone URL if the user may open a browser tab.`;
+        case "localhost":
+            return opts.standaloneUrl
+                ? `AGENT: Post a clickable markdown link to the user (e.g. "[Open ${label}](${opts.standaloneUrl})") — do not bury it in tool output or use broken workspace file paths. Re-post after the job completes if status was running.`
+                : `AGENT: Viz server unavailable — use jobs(status) and results(get) for progress and figures.`;
+        case "inline_image":
+            return `AGENT: Inline raster figure(s) are attached below. Summarize metrics in your reply; offer results_explorer or results(get) for additional figures.`;
+        case "text_only":
+            return `AGENT: No embedded panel or localhost viz — use jobs(status) for training progress and results(get) or results_explorer for figures.`;
+    }
+}
+export function resolveUiDelivery(options) {
+    const override = parseUiDeliveryOverride();
+    const mcpApps = getClientSupportsMcpApps();
+    const port = getVizPort() || null;
+    const standaloneUrl = buildStandaloneVizUrl(options.tool, options.jobId);
+    let delivery;
+    if (override === "apps") {
+        delivery = mcpApps ? "embedded" : port ? "localhost" : "text_only";
+    }
+    else if (override === "localhost") {
+        delivery = port ? "localhost" : options.inlineImageAttached ? "inline_image" : "text_only";
+    }
+    else if (override === "inline") {
+        delivery = options.inlineImageAttached ? "inline_image" : "text_only";
+    }
+    else {
+        // auto
+        if (mcpApps) {
+            delivery = "embedded";
+        }
+        else if (port) {
+            delivery = "localhost";
+        }
+        else if (options.inlineImageAttached) {
+            delivery = "inline_image";
+        }
+        else {
+            delivery = "text_only";
+        }
+    }
+    const urls = [];
+    if (standaloneUrl && (delivery === "embedded" || delivery === "localhost")) {
+        urls.push(standaloneUrl);
+    }
+    const healthUrl = buildHealthUrl(options.jobId);
+    if (healthUrl && delivery === "localhost") {
+        urls.push(healthUrl);
+    }
+    const crossLink = options.tool === "training_monitor" && options.jobStatus === "completed"
+        ? buildStandaloneVizUrl("results_explorer", options.jobId)
+        : options.tool === "results_explorer"
+            ? buildStandaloneVizUrl("training_monitor", options.jobId)
+            : undefined;
+    if (crossLink && !urls.includes(crossLink)) {
+        urls.push(crossLink);
+    }
+    return {
+        delivery,
+        urls: urls.length > 0 ? urls : undefined,
+        hint_for_agent: hintForTier(delivery, options.tool, { standaloneUrl, jobStatus: options.jobStatus }),
+        mcp_apps_supported: mcpApps,
+        viz_port: port,
+        port_pinned: vizPortPinned(),
+        override,
+    };
+}
+/**
+ * Prepend prominent standalone URL (hosts may advertise Apps but not mount them),
+ * then structured ui_delivery JSON + agent hint.
+ */
+export function appendUiDeliveryContent(content, block, options) {
+    const primary = options?.primaryUrl ?? block.urls?.[0];
+    const label = options?.linkLabel ?? "Open interactive view";
+    const prefix = [];
+    if (primary) {
+        const embedNote = block.delivery === "embedded"
+            ? "Embedded panel if your host supports MCP Apps — otherwise open the standalone URL below (expected on some Cursor builds)."
+            : "Open the standalone URL below when no inline panel appears.";
+        prefix.push({
+            type: "text",
+            text: `${embedNote}\n` +
+                `Standalone URL (copy if Open is blocked):\n${primary}\n` +
+                `[${label}](${primary})\n` +
+                `${block.hint_for_agent}`,
+        });
+    }
+    prefix.push({
+        type: "text",
+        text: "```json\n" + JSON.stringify({ ui_delivery: block }, null, 2) + "\n```",
+    });
+    if (primary && !block.port_pinned && block.delivery !== "text_only") {
+        prefix.push({
+            type: "text",
+            text: "Tip: set BARIVIA_VIZ_PORT in MCP env for a session-stable localhost port; re-run this tool if the page stops updating.",
+        });
+    }
+    content.unshift(...prefix);
+}
+export async function tryAttachLearningCurve(content, jobId) {
+    for (const ext of ["png", "svg"]) {
+        try {
+            const filename = `learning_curve.${ext}`;
+            const { data: buf } = await apiRawCall(`/v1/results/${jobId}/image/${filename}`);
+            content.push({
+                type: "image",
+                data: buf.toString("base64"),
+                mimeType: mimeForFilename(filename),
+                annotations: { audience: ["user"], priority: 0.75 },
+            });
+            return true;
+        }
+        catch {
+            continue;
+        }
+    }
+    return false;
+}
+export function getMcpClientDiagnostics() {
+    const override = parseUiDeliveryOverride();
+    const mcpApps = getClientSupportsMcpApps();
+    const port = getVizPort() || null;
+    let activeTier = "text_only";
+    if (override === "apps") {
+        activeTier = mcpApps ? "embedded" : port ? "localhost" : "text_only";
+    }
+    else if (override === "localhost") {
+        activeTier = port ? "localhost" : "text_only";
+    }
+    else if (override === "inline") {
+        activeTier = "inline_image";
+    }
+    else if (mcpApps) {
+        activeTier = "embedded";
+    }
+    else if (port) {
+        activeTier = "localhost";
+    }
+    return {
+        mcp_apps_supported: mcpApps,
+        viz_port: port,
+        viz_port_pinned: vizPortPinned(),
+        ui_delivery_override: override,
+        active_delivery_tier: activeTier,
+        health_url: buildHealthUrl(),
+    };
+}
+/** Reset inline budget at tool entry when attaching fallback images. */
+export function beginInlineFallback() {
+    resetInlineAttachBudget();
+}